Measurement of fermentation parameters of Chinese rice wine using Raman spectroscopy combined with linear and non-linear regression methods

Effective fermentation monitoring is a growing need during the manufacture of wine due to the rapid pace of change in the wine industry. Ethanol and reducing sugar are two most important process variables indicating the status of Chinese rice wine (CRW) fermentation process. In this study, the potentials of Raman spectroscopy (RS) as a rapid process analytical technique to monitor the evolution of these two chemical parameters involved in CRW fermentation process and to group samples according to different fermentation stages were investigated. The results demonstrated that compared with the PLS model using all wavelengths of Raman spectra, the prediction precision of model based on the spectral variables selected by competitive adaptive reweighted sampling (Cars) was significantly improved. In addition, nonlinear models outperformed linear models in predicting fermentation parameters. After systemically comparison and discussion, it was found that for both ethanol and glucose, Cars-support vector machine (Cars-SVM) models gave the best results with the highest prediction precisions. Moreover, the results obtained from discriminant partial least squares analysis (DPLS) showed that good performances were obtained with an average correct classification rate of 94.9% for different fermentation stages. The overall results indicated that RS combined with efficient variable selection algorithm and nonlinear regression tool could be utilized as a rapid method to monitor CRW fermentation process.     

支持向量机及其在石油勘探开发中的应用综述

支持向量机（SVM）是20世纪90年代中期发展起来的机器学习技术，与传统的人工神经网络不同，前者基于结构风险最小化原理，后者基于经验风险最小化原理。SVM在解决小样本、非线性和高维的机器学习问题中表现出许多特有的优势，适用于函数预测，模式识别和数据分类领域。将SVM应用于石油勘探开发领域是一个重要的研究方向，具有广阔的应用前景。     

A Support Vector Machine Approach for the Prediction of Drilling Fluid Density at High Temperature and High Pressure

Abstract

A support vector machine (SVM) approach was presented for predicting the drilling fluid density at high temperature and high pressure (HTHP). It is a universal model for water-based, oil-based, and synthetic drilling fluids. Available experimental data in the literature were used to develop and test this SVM model. Good agreement between SVM predictions and measured drilling fluid density values confirmed that the developed SVM model had good predictive precision and extrapolative features. The SVM model was also compared with the most popular models such as the artificial neural network (ANN) model, empirical correlations, and analytical models. Results showed that the SVM approach outperformed the competing methods for the prediction of drilling fluid density at HTHP.     

基于支持向量机方法的烷烃辛烷值预测

基于定量结构-性质相关(QSPR)原理,研究化学物质的结构与性能之间的关系,应用支持向量机(SVM)回归方法,建立了根据分子结构预测烷烃马达法辛烷值的数学模型,分别采用内部和外部验证的方式对模型性能进行了验证。结果表明,模型具有较高的稳定性以及预测能力。该方法的提出为工程上提供了一种根据分子结构有效预测烷烃马达法辛烷值的新方法。     

Prediction of total antioxidant capacity of red wine by Fourier transform infrared spectroscopy

Fourier transform infrared (FTIR) spectroscopy provides rapid and nondestructive analysis of wine, with almost no sample preparation. Aim of this study was to use FTIR measurement for the prediction of red wine total antioxidant capacity (TAC). Partial-least squares (PLS) regression was chosen for the evaluation of FTIR spectra. Plot of the full cross-validated PLS predicted TAC values showed a good correlation (r = 0.85), the slope of 0.74 and the prediction error provided by the PLS model was consistent with the uncertainty derived from the reference method. In conclusion, FTIR spectroscopy is a promising technique to rapidly provide information on TAC of red wines and has a high potential to be implemented for the rapid screening of several TAC methods concurrently.     

A comparison of Raman and FT-IR spectroscopy for the prediction of meat spoilage

In this study, time series spectroscopic, microbiological and sensory analysis data were obtained from minced beef samples stored under different packaging conditions (aerobic and modified atmosphere packaging) at 5 °C. These data were analyzed using machine learning and evolutionary computing methods, including partial least square regression (PLS-R), genetic programming (GP), genetic algorithm (GA), artificial neural networks (ANNs) and support vector machines regression (SVR) including different kernel functions [i.e. linear (SVR_L), polynomial (SVR_P), radial basis (RBF) (SVR_R) and sigmoid functions (SVR_S)]. Models predictive of the microbiological load and sensory assessment were calculated using these methods and the relative performance compared. In general, it was observed that for both FT-IR and Raman calibration models, better predictions were obtained for TVC, LAB and Enterobacteriaceae, whilst the FT-IR models performed in general slightly better in predicting the microbial counts compared to the Raman models. Additionally, regarding the predictions of the microbial counts the multivariate methods (SVM, PLS) that had similar performances gave better predictions compared to the evolutionary ones (GA-GP, GA-ANN, GP). On the other hand, the GA-GP model performed better from the others in predicting the sensory scores using the FT-IR data, whilst the GA-ANN model performed better in predicting the sensory scores using the Raman data. The results of this study demonstrate for the first time that Raman spectroscopy as well as FT-IR spectroscopy can be used reliably and accurately for the rapid assessment of meat spoilage.     

Application of infrared spectroscopy and partial least-squares for modeling the correlation of bitumen dynamic shear rheological with temperature

This paper reported a method for determining the correlation of the physical properties and the chemical composition of bitumen. The double-log model of dynamic shear modulus with temperature was performed for linearization treatment to obtained the constants, which were served as the independent variables using in partial least-squares(PLS) regression. In addition, infrared spectra were processed by data pretreatment methods and served as dependent variables in PLS. The regression vector b of PLS made possible the observation of the spectral regions contributing in the modeling, which provided a basis for research of the correlation between physical properties and chemical composition of bitumen.     

2D-fluorescence and multivariate data analysis for monitoring of sourdough fermentation process

Sourdough is used for the manufacture of bakery products, especially rye bread and vital for the development of its typical flavor. Although the process of sourdough fermentation is known thousands of years, still it is not understood in detail. Despite, in modern bread fabrication quality requirements are very high and demand a consequent control not only for the final baking process, but also for the production of intermediates. Characteristic process variables like pH-value and the degree of acidity are typically measured off-line to receive information about the state of the fermentation process. A new approach for monitoring the actual process state is the employment of 2D-fluorescence spectroscopy. As a non-invasive, optical method it is widely used for monitoring various types of bioprocesses, e.g. yeast or bacterial cultivations. In this contribution the application of partial least squares (PLS) regression and principal component regression (PCR) models for prediction of process variables of rye sourdough fermentations are compared to an evaluation where principal component analysis (PCA) is combined with artificial neural networks (ANN) for prediction of pH-value and acidity. For the pH-value PLS regression proved as good as PCR models and the combination of PCA and ANN. The average percentage root mean square error of prediction (pRMSEP) was between 2.5 and 5.1%. For the prediction of the acidity level, the best results were obtained using PLS regression models (pRMSEP: 6.0–8.1%). Smoothing the noisy 2D-fluorescence spectra slightly decreased the errors about 0.6%. Predictions with data sets of varying dough yield and constant temperature led to about 2% better results than data sets with different temperatures and constant dough yield. These results indicate a higher sensitivity of the prediction quality with respect to varying temperature compared to varying dough yield.     

支持向量机在裂缝预测及含气性评价应用中的优越性

为了对比多元回归分析（MRA）、人工神经网络（ANN）和支持向量机（SVM）三种算法的应用效果，分别将其应用于两个实例：对南襄盆地泌阳凹陷安棚油田安1井和安2井7种测井资料的34个样本进行裂缝预测；对鄂尔多斯盆地塔巴庙地区致密砂岩储集层孔隙度、渗透率和含气饱和度的40个样本进行含气性评价。两个实例分析表明：①非线性算法（SVM和ANN）远比线性算法（MRA）优越；②SVM表现绝对的优越性（计算误差均为零、计算速度快），是迄今最佳的机器学习方法；③在实例1中ANN与SVM相比几乎是同等优越，但在实例2中ANN的精度比SVM低得多；④MRA计算速度快、具有ANN和SVM所不具备的能表达研究目标与相关地质因素之间亲疏关系的优点。因此，当描述一个研究目标与多个相关地质因素的复杂关系时，应提倡采用SVM，而MRA可作为辅助应用。图4表3参27     

Qualitative and quantitative analysis of peanut adulteration in almond powder samples using multi-elemental fingerprinting combined with multivariate data analysis methods

In this study, adulteration of almond powder samples with peanut was analyzed using multi-elemental fingerprinting based on inductively coupled plasma optical emission measurements (ICP-OES) combined with chemometric methods. The ability of multivariate data analysis approaches, such as principal component analysis (PCA) and principal component analysis-linear discriminant analysis (PCA-LDA), to achieve differentiation of samples and as partial least squares (PLS) and least squares support vector machine (LS-SVM), to quantify the adulteration based on the elemental contents has been investigated. Ten variables i.e. the contents of B, Na, Mg, K, Ca, Fe, Cu, Cu, Zn and Sr at μg g⁻¹ level, determined by ICP-OES were used. Different almond and peanut samples were then mixed at various ratios to obtain mixtures ranging from 95/5 to 5/95 w/w and PCA-LDA was applied to classify the almonds, peanuts and adulterated samples. This method was able to differentiate peanut and almond samples from the adulterated samples. PLS and LS-SVM models were developed to quantify the adulteration ratios of almond using a training set and the constructed models were evaluated using a validation set. The root mean squared error of prediction (RMSEP) and the coefficient of determination (R²) of the validation set for PLS and LS-SVM were 3.81, 0.986 and 1.66, 0.997, respectively, which demonstrates the superiority of the LS-SVM model. The results show that the combination of multi-elemental fingerprinting with multivariate data analysis methods can be applied as an effective and feasible method for testing almond adulteration.     

Exploration of microwave dielectric and near infrared spectroscopy with multivariate data analysis for fat content determination in ground beef

This study investigated using microwave dielectric and near infrared (NIR) spectroscopy for the determination of fat content in ground beef samples (n = 69) in a designed experiment. Multivariate data analysis (principal component analysis (PCA) and partial least squares (PLS) regression modelling) was used to explore the potential of these spectroscopic techniques over selected multiple frequency or wavelength ranges. Chemical reference data for fat and water content in ground beef were obtained using a nuclear magnetic resonance-based SMART Trac analyser. Best performance of PLS prediction models for fat content revealed a coefficient of determination in prediction (R²P) of 0.87 and a root mean square error of prediction (RMSEP) of 2.71% w/w for microwave spectroscopy; in a similar manner, R²P of 0.99 and RMSEP of 0.71% w/w were obtained for NIR spectroscopy. The influence of water content on fat content prediction by microwave spectroscopy was investigated by comparing the prediction performance of PLS regression models developed using a single Y-variable (PLS1; fat or water content) and using both Y-variables (PLS2; fat and water contents).     

基于SVM的注水机组状态预示技术研究

油田大型注水机组在连续运转过程中,由于其自身的因素以及受外界条件的干扰,其运行常处于非线性非平稳状态。在充分研究和比较多种设备状态预示方法的基础上,提出一种基于支持向量机(Support Vector Machine,SVM)的状态预测新方法。该方法应用最终预报误差(FinalPrediction Error,FPE)准则确定样本的嵌入维数。通过比较SVM预测模型与自回归预测模型的单步和多步预测结果,证明基于SVM的预测方法在较长区间内具有良好的预测效果。用SVM预测大庆油田旋转注水机组时域的振动烈度,取得了较好预测效果,证明该算法能有效提高预测精度。     

基于支持向量机算法的注水管道剩余寿命预测

支持向量机是在统计学习理论的基础上发展而来的一种新的模式识别方法, 在解决有限样本、非线性及高维模式识别问题中表现出许多特有的优势。鉴于此, 针对注水管道的使用寿命和腐蚀影响因素之间复杂的映射关系, 在注水管道的剩余寿命预测研究中引入基于统计学习理论的支持向量机算法。研究了胜利油田某实验区注水水质腐蚀的影响因素, 应用LibSVM软件建立了注水管道的剩余寿命预测模型, 从而提供了一种预测注水管道剩余寿命的新方法。实际应用结果表明, 用支持向量机算法预测注水管道剩余寿命在样本有限的情况下具有明显优势。     

基于支持向量机的调剖效果预测方法研究

在统计分析大量已实施的调剖效果基础上,提出了基于支持向量机的调剖效果预测方法。通过分析影响调剖效果因素,运用支持向量机理论建立调剖效果预测模型。检验结果表明,基于支持向量机的调剖效果预测方法准确率较高,科学可行,应用前景广泛。     

基于灰狼算法优化的支持向量机产能预测

针对常规的线性回归以及经验公式等油井初期产能预测方法应用范围有限、预测误差较大,并且难以表征初产在多因素影响下的非线性变化规律等问题,提出了基于机器学习算法的产能预测方法。以某特低渗油田为例,从地质、开发和工程3个方面,选择了影响初期产能的10种因素,采用皮尔逊相关关系分析了各因素之间的线性相关性,使用随机森林方法确定了初期产能的主控因素,首次采用灰狼算法（GWO）优化的支持向量机（SVM）建立了油井初期产能的预测模型。结果表明：特低渗油田初期产能的主控因素为：压裂加砂量,射孔段厚度,初始含水饱和度,油层有效厚度和加砂强度;与多元线性回归模型和网格寻优的支持向量机模型相比,灰狼算法优化的支持向量机初期产能预测模型精度高而且运算速度快。研究结果可为油井初期产能评估提供参考。     

支持向量机的发展与应用

基于统计学习理论的支持向量机（SVM）是一种新型的机器学习方法，描述了SVM在模式识别和回归估计中的基本思想。在大训练样本情况下，用传统的方法求解SVM问题计算复杂，针对该问题探讨了一系列的SVM训练算法，并对其进行了比较。SVM由于其良好的泛化能力和全局最优性能．在模式识别、数据挖掘、非线性系统建模和控制等领域中展现出广泛的应用前景。     

支持向量机方法在油气储层参数预测中的应用

在油气储层综合研究中，储层参数的准确求取是一项关键性技术，而储层参数和地震信息之间并不存在直接的解析关系，不能用确定的函数表达式进行描述，通常采用数学统计的方法进行储层参数的预测。针对非线性函数拟合方法存在的困难，从Fourier多项式逼近的角度对非线性函数拟合的支持向量机的计算公式进行了的分析，其结论对理解和构造核函数提供了理论依据。支持向量机方法能够解决小样本情况下非线性函数拟合的通用性和推广性的问题，是求复杂的非线性拟合函数的一种非常有效的技术。模型及实例表明，该方法对油气储层参数的预测是有效的。     

基于核函数的非线性校正算法在原油快评中的应用

为了快速准确地测量原油的密度、酸值和硫质量分数等重要性质,采用红外光谱技术结合非线性化学计量学定量校正算法建立校正模型。结果表明,分别使用最小二乘支持向量机算法(LSSVM)和核偏最小二乘(KPLS)两种基于核函数的非线性校正算法建模预测原油密度、酸值和硫质量分数的预测标准偏差分别为00065 g/cm3、019 mgKOH/g和038%以及00089 g/cm3、023 mgKOH/g和040%,预测结果的重复性与再现性等同或优于标准方法。与经典偏最小二乘(PLS)方法相比,KPLS算法准确性更高,而LSSVM具有更快的训练速率、更小的测量偏差等优点。     

基于特征选择、遗传算法和支持向量机的水淹层识别方法

支持向量机是识别水淹层的有效方法,但其预测性能受多种因素的影响。研究提出一种水淹层识别新方法,采用Relief-F算法进行自动化特征选择,通过遗传算法优化模型参数以及使用加权支持向量机改善样本类数据分布不平衡对分类准确率的影响。将该方法应用于克拉玛依油田六中区克下组砾岩油藏水淹级别划分中,结果表明效果良好,增强了支持向量机的预测能力,进一步提高了水淹层解释的精度。     

基于PLS-MI组合的天牛须搜索BP神经网络模型对汽油辛烷值的预测性能

针对数据集中特征变量存在高度非线性和冗余的特点,提出了一种基于偏最小二乘回归(PLS)和互信息(MI)组合降维法的改进天牛须搜索算法（RSBAS）优化BP神经网络模型(PLS-MI-RSBASBP),并用于S Zorb脱硫装置汽油辛烷值的预测。首先通过偏最小二乘法和互信息组合算法选取与汽油辛烷值强相关的特征变量,然后使用RSBASBP模型对汽油辛烷值进行预测,并与BP,GABP,BASBP网络模型预测结果比较。结果表明：PLS-MI-RSBASBP模型预测结果较其他模型预测结果的MAE,MSE,RMSE更小,预测准确度高;而且,PLS-MI-RSBASBP模型可以确定影响汽油辛烷值的特征变量,从而进行有效控制和优化。